18 lines
606 B
Markdown
18 lines
606 B
Markdown
# VACEFlowAnnotator
|
||
Compute forward/backward optical flow between frames using VACE's flow model to capture motion cues.
|
||
|
||
## Inputs
|
||
- `images` – IMAGE batch with ≥2 frames.
|
||
- `flow_direction` (forward/backward/bidirectional) and `resolution`.
|
||
- Optional `model_path`.
|
||
|
||
## Outputs
|
||
- `flow_maps` – IMAGE tensor encoding motion vectors.
|
||
|
||
## Usage Tips
|
||
1. Only feed batches produced by VHS loaders or custom frame stacks.
|
||
2. Use bidirectional mode when you need symmetric motion guidance.
|
||
|
||
## Additional Resources
|
||
- [Full documentation](docs/nodes/vace-annotators/VACE_ANNOTATORS.md#vace-annotator-flow)
|